The great thing about OpenSCAD is that it makes creating models which may be described mathematically/programmatically easy.

The awful thing about OpenSCAD is that what one can create in it is strongly bounded by one's facility with and fluency in mathematics.

I'm curious how the new Python-enabled (I really hope it gets added in to the program itself) will affect this.

The awful thing is that you can only export triangles, so you can't make any true curves. If it was capable to true circles it would do everything I have ever needed to make.

OpenSCAD has it's origins in 3d printing (specifically the reprap project IIRC). Most 3d printing slicers[0] and firmware don't do curves[1].

[0] Some will import them now but convert them to triangles, cura can use a plugin to convert those triangles back to G2-G3 arcs or circles.

[1] at least by default, marlin has optional support for arcs that can be enabled at compile time.

I don't understand this. You can make triangles small enough that you can't tell the difference, and define them mathematically in terms of curves..? Isn't this how any rendering will ultimately work? The primitive is triangles.

Yes, but the representation as arcs allows one to defer the conversion to polylines/triangles until a later time.

As an example, if doing V carving, the appearance will be smoother if the geometry is imported so as to have arcs/curves, deferring the conversion to polylines/straight line movements until the last possible moment on the machine.

Arcs/Curves are also easier to work with if one needs to edit, and are more efficient in terms of representation/disk space/storage.

All true. Great comment.

Well, you could export enough triangles to get less than 0.0005" variance, but in practice actual CAM software will choke on it is my impression; I sent some circular objects out to be machined, but they just treated by OpenSCAD-exported files as a secondary reference for them redoing it in real software.

There is a big world out there beyond rendering.

CAM people, for example drive numerically controlled machines and robots. (CNC machines) These machines are driven by lines, arcs, higher order analytic curves, such as the conic, and recent controllers will accept bsplines.

Having path planning done with real geometric entities brings a lot of benefits along for the ride! Accuracy without a bazillion entities for the look ahead system to manage and for program buffers to drip feed (DNC) to a machine that probably does not have a large enough program memory to deal with that bazillion entities.

In many cases, derivative or extrapolated curves do not convey the right shape or path or are difficult and time consuming to force conformance.

We ended up with the big geometry kernels, such as Parasolid, precisely so that CAD could create a brep (boundary representation model) of sufficient fidelity to make other tasks, such as CAM and simulation, as well as derivative designs possible, practical, repeatable.

Those big code bodies have so damn many man hours of development in them it is very hard to replace it all. And that means we are stuck with some things, like many operations being single threaded, which caps CAD design to be a function of sequential compute speeds.

There are ways to build models to eliminate some dependencies, and the same goes for assemblies, so that multi-processing of some sort is possible. That brings our multi-core CPUs into the game.

But those are hard to teach, a lot of software does not yet support multi-threading anywhere near enough to fully exploit the model work, and, and, and...

In the CAD / CAM world, the primitive is the b-spline, and the surfaces it can represent.

Those generate faces, trim boundaries and sets of those create boundary representation models.

And still those can cause grief. They might be open, or just a surface representation. Or, they may self intersect...

Basically, every edge, which boils down to a b-spline curve that matches up to another b-spline curve serving as a trim boundary of a surface, must be shared by two and only two surfaces or the model is not representative of a real world, manufacturable thing.

Another great observation regarding real world objects is there are no edges, just radii, if only a single molecule big. Just an interesting and perhaps useful aside.

And b-splines are constrained to conform to the properties of lower order entities, such as lines, arcs, and the like. This is a part of why wireframe entities are all called "curves" or "sketch curves." They literally are all b-splines constrained by number and or placement of control points.

>You can make triangles small enough that you can't tell the difference, and define them mathematically in terms of curves..?

This doesn't work if you already have a mesh. You need to generate your triangles on demand from the BREP data.

Alternatively you need to use tesselation, but then you need to design your models with tesselation in mind, such as people do in blender.

>Isn't this how any rendering will ultimately work? The primitive is triangles.

Uhm no. 3D printing and CAM do not work with triangles. The slicer and CAM software generate tool paths in gcode.

Importing STEP files into OrcaSlicer/BambuSlicer gives me better quality results and better print speeds than using STL files.

I have personally never understood the love for STL files in the 3D printing community for anything that isn't the result of a blender export. Like, is it really that fun to constantly export different versions of STL files set to different quality levels? Rescaling a print in the slicer is not recommended with STL. You have to re-export the model from blender where the tesselation happens.

I added the ability to export to arcs/circles when writing out DXFs using PythonSCAD in the library:

https://github.com/WillAdams/gcodepreview

See:

https://github.com/WillAdams/gcodepreview/blob/main/export.d...

You call out 2d and mention mills and routers, so after reading the page I'm thinking I wouldn't be able to design a piece to be put in a CNC lathe ala the hub in the lower right of https://canev.com/cdn/shop/products/[email protected]..., which has multiple holes (sometimes threaded), including threaded setscrew holes tangential to the primary axis.

Not as-of-yet --- it's 3-axis at the moment --- hoping to eventually add support for additional axis/axes.

Have you looked at grblgru.com ? Not opensource, but does have CNC lathe support.

It's not a serious problem I have, just saying I like OpenSCAD, I have been able to generate every shape I've ever needed to create, and it would do everything I want if it was possible to generate true curves (because I farm out the manufacturing).

Do you have a sample DXF (which uses arcs) which describes a lathe project?

I dug up an OpenSCAD from some backups if that's interesting to you https://pastebin.com/xwhrLwQH

I do not have any DXF of something like this.

You said this so much more succinctly than me.

> People do get oddly grumpy when you criticise OpenSCAD

Most of us don't realize how much our comfort with familiar concepts affects our ability to objectively evaluate them. Moving from dev to being a tech artist, I see the same exact sort of irrational biases, hang-ups and endemic misconceptions with completely different tool chains and types of complexity. When I first started using Houdini, I tried to turn every task into a Python coding problem because that's what I was comfortable with— Python is amazing in Houdini but it's definitely not the best way to accomplish most tasks.

This viewpoint resonates for me. OpenSCAD is a useful tool for hacking together 3D parts quickly, but an excellent tool for exercising the shape rotating parts of your brain.

It's also good if your aim is to visualise maths.

There is actually a book on geometry which takes advantage of this:

https://www.goodreads.com/book/show/58059196-make

which has since become a series including:

Trigonometry: https://www.goodreads.com/book/show/123127774-make

Calculus: https://www.goodreads.com/book/show/61739368-make

Currently working through a library using the new Python-enabled OpenPythonSCAD, and at the end of it, hope to work on a project which requires conic sections (which I'm still researching, so if anyone has any good resources on that I'd be glad of them).

Yes - I haven't bought these yet but I do intend to, because I need to brush up on all three (well, I don't think I ever really learned calculus, to be honest).

Joan Horvath is a superhero -- I've seen a few interviews now.

I will very happily use OpenSCAD for that because it seems like a well-chosen tool for the job.

I think that sounds like a healthy approach to social ad platforms and the like.